Forecasting Credit Portfolio Risk - Discussion Paper Series 2: Banking and Financial Supervision No 01/2004 pot

Forecasting Credit Portfolio Risk Alfred Hamerle (Universität Regensburg) Thilo Liebig (Deutsche Bundesbank) Harald Scheule (Universität Regensburg) Discussion Paper Series 2: Banking and Financial Supervision No 01/2004 Discussion Papers represent the authors’ personal opinions and not necessarily reflect the views of the Deutsche Bundesbank or its staff Editorial Board: Heinz Herrmann Thilo Liebig Karl-Heinz Tödter Deutsche Bundesbank, Wilhelm-Epstein-Strasse 14, 60431 Frankfurt am Main, Postfach 10 06 02, 60006 Frankfurt am Main Tel +49 69 9566-1 Telex within Germany 41227, telex from abroad 414431, fax +49 69 5601071 Please address all orders in writing to: Deutsche Bundesbank, Press and Public Relations Division, at the above address or via fax No +49 69 9566-3077 Reproduction permitted only if source is stated ISBN 3–935821–82–4 Abstract The main challenge of forecasting credit default risk in loan portfolios is forecasting the default probabilities and the default correlations We derive a Merton-style threshold-value model for the default probability which treats the asset value of a firm as unknown and uses a factor model instead In addition, we demonstrate how default correlations can be easily modeled The empirical analysis is based on a large data set of German firms provided by Deutsche Bundesbank We find that the inclusion of variables which are correlated with the business cycle improves the forecasts of default probabilities Asset and default correlations depend on the factors used to model default probabilities The better the point-in-time calibration of the estimated default probabilities, the smaller the estimated correlations Thus, correlations and default probabilities should always be estimated simultaneously Keywords: asset correlation, bank regulation, Basel II, credit risk, default correlation, default probability, logit model, probit model, time-discrete hazard rate JEL classification: C23, C41, G21 Non-technical Summary Forecasting credit portfolio risk poses a challenge for the banking industry One important goal of modern credit portfolio models is the forecast of the future credit risk given the information which is available at the point of time the forecast is made Thus, the discussion paper “Forecasting Credit Portfolio Risk“ proposes a dynamic concept for the forecast of the risk parameters default probabilities and default correlations The results are based on an extensive empirical analysis of a data set provided by Deutsche Bundesbank which contains financial statements for more than 50,000 German firms and a time period from 1987 to 2000 Important results of this paper are: The inclusion of macroeconomic risk drivers improves the forecast of default probabilities considerably We included the macroeconomic variables business climate index, unemployment rate and systematic growth in new orders of the construction industry We find that a large part of co-movements can be attributed to lagged risk drivers Thus, default rate or loss distributions can be forecasted given the values of the lagged risk drivers The model allows default probabilities to be forecasted for individual borrowers and to estimate correlations between those borrowers simultaneously We show that asset and default correlations depend on the point in time calibration of the default probabilities In addition a simultaneous estimation eases the validation of default probabilities Thus, default probabilities and correlations should never be derived separately from each other The model is an empirical application of the model which is used for the calibration of risk weights by the Basel Committee on Banking Supervision Hence, we are able to compare the estimated parameters from our model and Basel II directly Nichttechnische Zusammenfassung Die Prognose von Kreditausfallrisiken stellt eine zentrale Herausforderung für Kreditinstitute und Finanzdienstleister dar Ein wichtiges Ziel moderner Kreditrisikomodelle ist die Prognose zukünftiger Kreditrisiken auf Basis der im Prognosezeitpunkt zur Verfügung stehenden Information Vor diesem Hintergrund präsentiert der Diskussionsbeitrag “Forecasting Credit Portfolio Risk“ ein dynamisches Konzept zur gemeinsamen Prognose der zentralen Risikoparameter Ausfallwahrscheinlichkeit und Ausfallkorrelation Die empirischen Untersuchungen in dieser Arbeit basieren auf der Unternehmensbilanzdatenbank der Deutschen Bundesbank Wichtige Ergebnisse des Diskussionsbeitrags sind: Die Berücksichtigung von makroökonomischen Einflgrưßen verbessert signifikant die Güte der Prognose von Ausfallwahrscheinlichkeiten Als makrkonomische Einflgrưßen wurden der Ifo-Geschäftsklimaindex, die Arbeitslosenquote und die Auftragseingänge der Baubranche verwendet Ausfallwahrscheinlichkeiten und Ausfallkorrelationen können durch zeitverzögert wirkende Risikofaktoren erklärt werden Resultierende Verlustverteilungen können deshalb bei Kenntnis der Ausprägungen der Risikofaktoren prognostiziert werden Der Modellansatz erlaubt erstmals die simultane Ermittlung von Ausfallwahrscheinlichkeiten und Ausfallkorrelationen Mit der Point-in-Time-Kalibrierung der Ausfallwahrscheinlichkeiten nehmen die geschätzten Korrelationen ab Des Weiteren erleichtert die simultane Schätzung die Validierung der Ausfallwahrscheinlichkeiten Korrelationen und Ausfallwahrscheinlichkeiten sollten deshalb nicht getrennt voneinander ermittelt werden Das Modell entspricht dem des Baseler Ausschusses für Bankenaufsicht Die geschätzten Parameter können deshalb unmittelbar mit den Basel II Vorgaben verglichen werden Content Introduction .1 Modeling default probabilities 3 Modeling correlations .8 Empirical Analysis 12 4.1 4.2 Model-estimation for one risk segment 14 4.3 Model-estimation for multiple risk segments 19 4.4 Forecasting default probabilities 23 4.5 Data 12 Forecasting the default rate distribution 24 Summary 27 Appendix 27 References 28 Forecasting Credit Portfolio Risk* Introduction The main challenge of forecasting credit default risk in loan portfolios is forecasting the default probabilities and the default correlations They are input parameters to a variety of credit risk models like CreditMetricsä, CreditRisk+, CreditPortfolioManagerä or CreditPortfolioViewä For outlines of these models see Gupton et al [1997], Credit Suisse Financial Products [1997], Crosbie/Bohn [2002] and Wilson [1997a, 1997b] The main direction of modeling credit risk has its origin in the seminal model of Merton [1974, 1977] and Black/Scholes [1973] Extensions of the approach are described in Black and Cox [1976], Merton [1977], Geske [1977], Longstaff and Schwartz [1995] or Zhou [2001] In this model it is assumed that a default event happens if the value of an obligor’s assets falls short of the value of debt Generally speaking, one of the model’s major shortcomings is the assumption of available market prices for the asset value This is not usually valid for retail or small and medium-sized obligors Chart displays West German insolvency rates for the years 1980 to 2000 Insolvency rates are frequently taken as proxies for default rates It can be seen that the rates fluctuate over time An important object of modern credit risk management is the forecast of future credit risk given the available information at the point of time at which the forecast is made * We would like to thank Dr Stefan Blochwitz, Dr Klaus Düllmann and Dr Daniel Rösch for stimulating discussions -1- 0.01 insolvency rate 0.008 0.006 0.004 0.002 1980 1985 1990 1995 2000 year Chart 1: Insolvency rates of West Germany In the present paper we use a model to forecast default probabilities and estimate default correlations based on the threshold model described above The default probability measures the probability of an obligor’s assets falling short of a threshold In addition, asset correlations are modeled as a measure of co-movement of the asset values of two obligors Default correlations can then be derived analytically Our approach differs from existing studies on forecasting default probabilities (such as Escott/ Glormann/ Kocagil [2001], Falkenstein [2000] and Shumway [2001]) and estimating default correlations (like Dietsch/ Petey [2002], Gupton/Finger/Bhatia [1997] and Lucas [1995]) in several ways and therefore leads to new important results Firstly, we find that a large part of co-movements can be attributed to lagged risk drivers Thus, default rate or loss distributions can be forecasted, given the values of the lagged risk drivers, and estimation uncertainty can be reduced Secondly, the model we employ allows default probabilities to be forecasted for -2- Risk segment Parameter estimate Standard error P-value Manufacturing 0.0796 0.0226 0.1167 Commerce 0.1234 0.0347 0.0495 Others 0.0990 0.0315 0.2033 Table 7: Random effect parameter estimates and significance, model The inclusion of the macroeconomic risk drivers results in a decrease of the estimated parameter of the random effect and therefore the asset correlation As a matter of fact, the random effect becomes insignificant ( α = 0,05 ) The asset correlation of the risk segment Commerce remains unchanged because no significant macroeconomic variable was found Note that the asset correlation of this segment in model is already lower than the ones of the other segments Table summarizes the asset correlations for obligors of the same and different risk segments for model and Table for model Manufacturing Manufacturing Commerce Others 0.0071 0.0040 0.0038 0.0046 0.0034 Commerce 0.0136 Others Table 8: Asset correlation estimates, model Manufacturing Manufacturing Commerce Others 0.0019 0.0015 0.0003 0.0046 0.0016 Commerce 0.0030 Others Table 9: Asset correlation estimates, model - 22 - 4.4 Forecasting default probabilities Time-varying variables enter the logit model with a time lag Thus, given the estimated models to and the value of the risk drivers, default probabilities will now be forecasted for the year 2000 Chart compares the empirical frequency distribution (class width: 0.001) of the forecasted default probabilities of model and model 2: 0.2 relative frequency 0.15 0.1 0.05 0 0.01 0.02 0.03 0.04 0.05 forecasted default probability model model Chart 7: Frequency distribution of forecasted default probabilities, model and model Chart compares the real and the mean forecasted default rates of model to model The forecasted default rate is the average of the forecasted default probabilities - 23 - 0.009 default rate 0.008 0.007 0.006 0.005 0.004 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 year real default rate forecasted default rate (model and 3) forecasted default rate (model and 4) Chart 8: Real and forecasted default rates for year 2000, model to model Model and model which include macroeconomic variables, forecast the default rate more accurately than model and model which not include macroeconomic variables Note that the forecasted default rate for model and are very close to each other and therefore can not be differentiated in the chart In other words, the calibration of the forecasted default probabilities would have been better if macroeconomic variables had been included in the respective model 4.5 Forecasting the default rate distribution The forecasted default probabilities and the estimated asset correlations can be aggregated to the forecasted default rate distribution The forecasted default rate distribution can be interpreted as a loss distribution if the exposure at default and the loss given default equal one - 24 - Chart compares the forecasted default rate distribution of model without macroeconomic variables and model with macroeconomic variables Table 10 shows the respective mean forecasted default rate and the quantiles of the forecasted default rate distribution forecasted probability 0.16 0.12 0.08 0.04 0 0.005 0.01 0.015 0.02 0.025 default rate model model real default rate Chart 9: Forecasted default rate distribution, model and model Mean forecasted default rate 0,95-Q 0,99-Q 0,999-Q Model 0.0134 0.0167 0.0187 0.0217 Model 0.0146 0.0170 0.0183 0.0201 Table 10: Mean forecasted default rate and quantiles of forecasted default rate distribution, model and model Again, it can be seen that the mean forecasted default rate for 2000 of model is closer to the real default rate than that of model In addition, model estimates a lower asset correlation - 25 - which leads to a lower variance of the forecasted default rate Hence, the portfolio credit risk is forecasted more accurately Similar results are observed for model and model when multiple risk segments are assumed Chart 10 compares the forecasted default rate distribution of one risk segment model and multiple risk segment model Table 11 shows the respective mean forecasted default rate and the quantiles of the forecasted default rate distribution: forecasted probability 0.16 0.12 0.08 0.04 0 0.005 0.01 0.015 0.02 0.025 default rate model model real default rate Chart 10: Forecasted default rate distribution, model and model Mean forecasted default rate 0,95-Q 0,99-Q 0,999-Q Model 0.0134 0.0167 0.0187 0.0217 Model 0.0131 0.0163 0.0181 0.0206 Table 11: Mean forecasted default rate and quantiles of forecasted default rate distribution, model and model - 26 - The forecasted default rate distribution of model is broader than that of model Since default rate distributions generally broaden with a higher mean forecasted default rate, we cannot conclude that the assumption of multiple risk segments leads to more accurate credit portfolio risk forecasts An examination of further periods is advisable Summary The present paper describes an alternative methodology for forecasting credit portfolio risk We showed within this framework that • individual default probabilities can be forecasted and asset (or default) correlations can be estimated, given the values of risk drivers that are observable in the point of time the forecasts are made • the inclusion of variables which are correlated with the business cycle improves the forecasts of default probabilities The variance of the forecasted default rate decreases, i.e the uncertainty of the forecasts is diminished • asset and default correlations depend on the factors used to model default probabilities The better the point-in-time calibration of the estimated default probabilities, the smaller the estimated correlations Thus, correlations and default probabilities should always be estimated simultaneously - 27 - Appendix • Descriptive statistics firm-specific risk drivers Ratio Mean Median Standarddev Min Max ART 34.5973 30.7526 25.8769 100 APT 34.8201 25.9981 30.1343 110 CFT 6.5248 4.7671 9.3863 -15 25 CRR 12.2592 10.0036 16.2009 -25 50 ETA 12.1587 9.7701 20.6234 -35 60 ITT 48.1980 38.0972 44.0308 160 RIE 387.7100 218.4480 486.0730 -650 1,400 TTT 0.0076 0.0074 0.0011 0.0062 0.0096 Table 12: Data set Deutsche Bundesbank - summary statistics of firm-specific risk drivers ART ART APT CFT CRR APT CFT CRR ETA ITT RIE TTT 1.0000 0.2212 -0.1255 -0.0949 -0.0400 0.0860 -0.0333 0.0675 1.0000 -0.0757 -0.1780 -0.2418 0.2460 -0.2225 -0.0584 1.0000 0.7423 0.1812 -0.2627 0.3303 -0.1666 1.0000 0.1476 -0.2923 0.5290 -0.0227 1.0000 -0.0351 0.3423 -0.0588 1.0000 -0.1924 0.0121 1.0000 -0.0174 ETA ITT RIE 1.0000 TTT Table 13: Data set Deutsche Bundesbank - Pearson correlations between firm-specific risk drivers - 28 - • Descriptive statistics macroeconomic risk drivers Variable Mean Median Standarddev Min Max BCI 88.311 85.028 7.483 82.29 103.36 GOC 0.0013 0.0054 0.0206 -0.0270 0.0369 Insolvency rate 0.0072 0.0073 0.0013 0.0050 0.0084 UER 9.0537 9.3259 1.8706 6.3000 11.4830 Table 14: Data set Deutsche Bundesbank - summary statistics of macroeconomic risk drivers BCI BCI GOC GOC Insolvency Rate UER 1.0000 0.7734 -0.8278 -0.7685 1.0000 -0.7851 -0.7662 1.0000 0.6347 Insolvency Rate 1.0000 UER Table 15: Data set Deutsche Bundesbank - Pearson correlations between macroeconomic risk drivers - 29 - References Basel Committee on Banking Supervision, The New Basel Capital Accord, Consultative Document, April 2003, Bank for International Settlement, Basel Black, F./ Cox, J.C., Valuing Corporate Securities: Some Effects of Bond Indenture Provisions, 1976, Journal of Finance, 31, pp 351- 367 Black, F./ Scholes, M., The Pricing of Options and Corporate Liabilities, 1973, Journal of Political Economy, 81, pp 637- 654 Credit Suisse Financial Products, CreditRisk+ - A Credit Risk Management Framework, 1997, London Crosbie, P.J./ Bohn, J.R., Modeling Default Risk, 2002, KMV, San Francisco Davidson, R./ MacKinnon, J.G., Estimation and Inference in Econometrics, 1993, Oxford University Press, New York Dietsch, M./ Petey, J., The credit risk in SME loans portfolios: Modeling issues, pricing, and capital requirements, 2002, Journal of Banking and Finance, 26, pp 303- 322 Engelmann, B./ Hayden, E./ Tasche, D., Testing for Rating Accuracy, 2003, Risk, 16, January, pp 82-86 Escott, P./ Glormann, F./ Kocagil, A.E., Moody’s RiskCalcä for Private Companies: the German Model, 2001, Moody’s Investors Service, New York Falkenstein, E., RiskCalcä for Private Companies: Moody’s Default Model, 2000, Moody’s Investors Service, New York Finger, C.C., The One-Factor CreditMetrics Model in the New Basle Capital Accord, 2001, RiskMetrics Journal, pp 9- 18 Geske, R., The Valuation of Corporate Liabilities as Compound Options, 1977, Journal of Financial and Quantitative Analysis, 12, pp 541- 552 - 30 - Gupton, G.M./ Finger, C.C./ Bhatia, M., CreditMetrics Technical Document, 1997, J.P Morgan & Co., New York Longstaff, F.A./ Schwartz, E.S., A Simple Approach to Valuing Risky Fixed and Floating Rate Debt, 1995, Journal of Finance, 50, pp 789- 819 Lucas, D.J., Default Correlation and Credit Analysis, 1995, Journal of Fixed Income, pp 76- 87 Merton, R C., On the Pricing of Corporate Debt: The Risk Structure of Interest Rates, 1974, Journal of Finance, 29, pp 449- 470 Merton, R C., On the Pricing of Contingent Claims and the Modigliani-Miller Theorem, 1977 Journal of Financial Economics, 5, pp 241- 249 Pinheiro, J.C./ Bates, D.M., Approximations to the Log-Likelihood Function in the Nonlinear Mixed-Effects Model, 1995, Journal of Computational and Graphical Statistics, 4, pp 12- 35 Rabe-Hesketh, S./ Skrondal, A./ Pickles, A., Estimation of generalized linear mixed models, 2002, Stata Journal, 2, pp 1- 21 Scheule, H., Prognose von Kreditausfallrisiken, 2003, Dissertation, University of Regensburg, Uhlenbruch Verlag Shumway, T., Forecasting Bankruptcy More Accurately: A Simple Hazard Model, 2001, The Journal of Business, 74, pp 101- 124 Sobehart, J.R./ Keenan, S.C./ Stein, R.M., Benchmarking Quantitative Default Risk Models: A Validation Methodology, 2000, Moody’s Investors Service, New York Wilson, T.C., Portfolio Credit Risk I, 1997a, Risk, 10, September, pp 111- 117 Wilson, T.C., Portfolio Credit Risk II, 1997b, Risk, 10, October, pp 56- 61 Zhou, C., An Analysis of Default Correlations and Multiple Defaults, 2001, Review of Financial Studies, 14, pp 555- 576 - 31 - The following Discussion Papers have been published since 2003: Series 1: Studies of the Economic Research Centre January 2003 Testing mean-variance efficiency in CAPM with possibly non-gaussian errors: an exact simulation-based approach Marie-Claude Beaul Jean-Marie Dufour Lynda Khalaf January 2003 Finite-sample distributions of self-normalized sums Jeong-Ryeol Kim 2003 The stock return-inflation puzzle and the asymmetric causality in stock returns, inflation and real activity Jeong-Ryeol Kim January February 2003 Multiple equilibrium overnight rates in a dynamic interbank market game Jens Tapking February 2003 A comparison of dynamic panel data estimators: Monte Carlo evidence and an application to the investment function Andreas Behr March 2003 A Vectorautoregressive Investment Model (VIM) And Monetary Policy Transmission: Panel Evidence From German Firms March March Joerg Breitung Robert S Chirinko Ulf von Kalckreuth 2003 The international integration of money markets in the central and east European accession countries: deviations from covered interest parity, capital controls and inefficienCies in the financial sector Sabine Herrmann Axel Jochem 2003 The international integration of foreign exchange markets in the central and east European accession countries: speculative efficiency, transaction costs and exchange rate premiums Sabine Herrmann Axel Jochem - 33 - March 2003 Determinants of German FDI: New Evidence from Micro-Data Claudia Buch Jörn Kleinert Farid Toubal March 2003 On the Stability of Different Financial Systems Falko Fecht Determinants of German Foreign Direct Investment in Latin American and Asian Emerging Markets in the 1990s Torsten Wezel Active monetary policy, passive fiscal policy and the value of public debt: some further monetarist arithmetic Leopold von Thadden April 2003 June June 2003 2003 Bidder Behavior in Repo Auctions without Minimum Bid Rate:Dieter Nautz Evidence from the Bundesbank Tobias Linzert Jörg Breitung June 2003 Did the Bundesbank React to Stock Price Movements? Martin T Bohl Pierre L Siklos Thomas Werner 15 2003 Money in a New-Keynesian model estimated with German data Jana Kremer Giovanni Lombardo Thomas Werner 16 2003 Exact tests and confidence sets for the tail coefficient of α-stable distributions Jean-Marie Dufour Jeong-Ryeol Kurz-Kim 17 2003 The Forecasting Performance of German Stock Option Densities B R Craig, E Glatzer, J Keller, M Scheicher 18 2003 How wacky is the DAX? The changing structure of German stock market volatility Jelena Stapf Thomas Werner - 34 - 2004 Foreign Bank Entry into Emerging Economies: An Empirical Assessment of the Determinants and Risks Predicated on German FDI Data Torsten Wezel 2004 Does Co-Financing by Multilateral Development Banks Increase “Risky” Direct Investment in Emerging Markets? – Evidence for German Banking FDI Torsten Wezel 2004 Policy Instrument Choice and Non-Coordinated Monetary Policy in Interdependent Economies Giovanni Lombardo Alan Sutherland 2004 Inflation Targeting Rules and Welfare in an Asymmetric Currency Area Giovanni Lombardo 2004 FDI versus cross-border financial services: The globalisation of German banks Claudia M Buch Alexander Lipponer 2004 Clustering or competition? The foreign investment behaviour of German banks Claudia M Buch Alexander Lipponer Series 2: Banking and Financial Supervision 2003 Measuring the Discriminative Power of Rating Systems B Engelmann, E Hayden, D Tasche 2003 Credit Risk Factor Modeling and the Basel II IRB Approach A Hamerle, T Liebig, D Rösch 2004 Forecasting Credit Portfolio Risk A Hamerle, T Liebig, H Scheule - 35 - ... 1.0000 -0 .0757 -0 .1780 -0 .2418 0.2460 -0 .2225 -0 .0584 1.0000 0.7423 0.1812 -0 .2627 0.3303 -0 .1666 1.0000 0.1476 -0 .2923 0.5290 -0 .0227 1.0000 -0 .0351 0.3423 -0 .0588 1.0000 -0 .1924 0.0121 1.0000 -0 .0174... models: Risk driver Model (without Model (with macroeconomic risk driver) macroeconomic risk driver) Intercept -7 .7832 -7 .8132 ART 0.0062 0.0062 APT 0.0123 0.0124 CRR -0 .0324 -0 .0326 ETA -0 .0162 -0 .0160... variety of credit risk models like CreditMetricsä, CreditRisk+, CreditPortfolioManagerä or CreditPortfolioViewä For outlines of these models see Gupton et al [1997], Credit Suisse Financial Products